ARLArena: A Unified Framework for Stable Agentic Reinforcement Learning

TL;DR

This paper introduces ARLArena, a systematic framework for analyzing and improving the stability of agentic reinforcement learning, leading to the development of SAMPO, a method that ensures stable training across various tasks.

Contribution

It presents ARLArena, a standardized testbed and analysis framework, and proposes SAMPO, a novel policy optimization method that enhances stability in agentic reinforcement learning.

Findings

SAMPO achieves stable training across diverse tasks.

ARLArena provides a systematic analysis of stability factors.

The framework offers practical guidance for LLM-based agent training.

Abstract

Agentic reinforcement learning (ARL) has rapidly gained attention as a promising paradigm for training agents to solve complex, multi-step interactive tasks. Despite encouraging early results, ARL remains highly unstable, often leading to training collapse. This instability limits scalability to larger environments and longer interaction horizons, and constrains systematic exploration of algorithmic design choices. In this paper, we first propose ARLArena, a stable training recipe and systematic analysis framework that examines training stability in a controlled and reproducible setting. ARLArena first constructs a clean and standardized testbed. Then, we decompose policy gradient into four core design dimensions and assess the performance and stability of each dimension. Through this fine-grained analysis, we distill a unified perspective on ARL and propose SAMPO, a stable agentic policy optimization method designed to mitigate the dominant sources of instability in ARL. Empirically, SAMPO achieves consistently stable training and strong performance across diverse agentic tasks. Overall, this study provides a unifying policy gradient perspective for ARL and offers practical guidance for building stable and reproducible LLM-based agent training pipelines.